Acute and chronic wounds are a challenge to healthcare systems around the world and affect many people's lives annually. Wound classification is a key step in wound diagnosis that would help clinicians to identify an optimal treatment procedure. Hence, having a high-performance classifier assists the specialists in the field to classify the wounds with less financial and time costs. Different machine learning and deep learning-based wound classification methods have been proposed in the literature. In this study, we have developed an ensemble Deep Convolutional Neural Network-based classifier to classify wound images including surgical, diabetic, and venous ulcers, into multi-classes. The output classification scores of two classifiers (patch-wise and image-wise) are fed into a Multi-Layer Perceptron to provide a superior classification performance. A 5-fold cross-validation approach is used to evaluate the proposed method. We obtained maximum and average classification accuracy values of 96.4% and 94.28% for binary and 91.9\% and 87.7\% for 3-class classification problems. The results show that our proposed method can be used effectively as a decision support system in classification of wound images or other related clinical applications.
Active learning strives to reduce annotation costs by choosing the most critical examples to label. Typically, the active learning strategy is contingent on the classification model. For instance, uncertainty sampling depends on poorly calibrated model confidence scores. In the cold-start setting, active learning is impractical because of model instability and data scarcity. Fortunately, modern NLP provides an additional source of information: pre-trained language models. The pre-training loss can find examples that surprise the model and should be labeled for efficient fine-tuning. Therefore, we treat the language modeling loss as a proxy for classification uncertainty. With BERT, we develop a simple strategy based on the masked language modeling loss that minimizes labeling costs for text classification. Compared to other baselines, our approach reaches higher accuracy within less sampling iterations and computation time.
Relation classification is one of the key topics in information extraction, which can be used to construct knowledge bases or to provide useful information for question answering. Current approaches for relation classification are mainly focused on the English language and require lots of training data with human annotations. Creating and annotating a large amount of training data for low-resource languages is impractical and expensive. To overcome this issue, we propose two cross-lingual relation classification models: a baseline model based on Multilingual BERT and a new multilingual pretraining setup, which significantly improves the baseline with distant supervision. For evaluation, we introduce a new public benchmark dataset for cross-lingual relation classification in English, French, German, Spanish, and Turkish, called RELX. We also provide the RELX-Distant dataset, which includes hundreds of thousands of sentences with relations from Wikipedia and Wikidata collected by distant supervision for these languages. Our code and data are available at: https://github.com/boun-tabi/RELX
This paper presents our submission to Task 2 of the Workshop on Noisy User-generated Text. We explore improving the performance of a pre-trained transformer-based language model fine-tuned for text classification through an ensemble implementation that makes use of corpus level information and a handcrafted feature. We test the effectiveness of including the aforementioned features in accommodating the challenges of a noisy data set centred on a specific subject outside the remit of the pre-training data. We show that inclusion of additional features can improve classification results and achieve a score within 2 points of the top performing team.
Supervised classification techniques use training samples to find classification rules with small expected 0-1 loss. Conventional methods achieve efficient learning and out-of-sample generalization by minimizing surrogate losses over specific families of rules. This paper presents minimax risk classifiers (MRCs) that do not rely on a choice of surrogate loss and family of rules. MRCs achieve efficient learning and out-of-sample generalization by minimizing worst-case expected 0-1 loss w.r.t. uncertainty sets that are defined by linear constraints and include the true underlying distribution. In addition, MRCs' learning stage provides performance guarantees as lower and upper tight bounds for expected 0-1 loss. We also present MRCs' finite-sample generalization bounds in terms of training size and smallest minimax risk, and show their competitive classification performance w.r.t. state-of-the-art techniques using benchmark datasets.
We propose several new models for semi-supervised nonnegative matrix factorization (SSNMF) and provide motivation for SSNMF models as maximum likelihood estimators given specific distributions of uncertainty. We present multiplicative updates training methods for each new model, and demonstrate the application of these models to classification, although they are flexible to other supervised learning tasks. We illustrate the promise of these models and training methods on both synthetic and real data, and achieve high classification accuracy on the 20 Newsgroups dataset.
It has been challenging to identify ferrograph images with a small dataset and various scales of wear particle. A novel model is proposed in this study to cope with these challenging problems. For the problem of insufficient samples, we first proposed a data augmentation algorithm based on the permutation of image patches. Then, an auxiliary loss function of image patch permutation recognition was proposed to identify the image generated by the data augmentation algorithm. Moreover, we designed a feature extraction loss function to force the proposed model to extract more abundant features and to reduce redundant representations. As for the challenge of large change range of wear particle size, we proposed a multi-scale feature extraction block to obtain the multi-scale representations of wear particles. We carried out experiments on a ferrograph image dataset and a mini-CIFAR-10 dataset. Experimental results show that the proposed model can improve the accuracy of the two datasets by 9% and 20% respectively compared with the baseline.
Cyberbullying is a prevalent social problem that inflicts detrimental consequences to the health and safety of victims such as psychological distress, anti-social behaviour, and suicide. The automation of cyberbullying detection is a recent but widely researched problem, with current research having a strong focus on a binary classification of bullying versus non-bullying. This paper proposes a novel approach to enhancing cyberbullying detection through role modeling. We utilise a dataset from ASKfm to perform multi-class classification to detect participant roles (e.g. victim, harasser). Our preliminary results demonstrate promising performance including 0.83 and 0.76 of F1-score for cyberbullying and role classification respectively, outperforming baselines.
The decision tree is one of the most popular and classical machine learning models from the 1980s. However, in many practical applications, decision trees tend to generate decision paths with excessive depth. Long decision paths often cause overfitting problems, and make models difficult to interpret. With longer decision paths, inference is also more likely to fail when the data contain missing values. In this work, we propose a new tree model called Cascading Decision Trees to alleviate this problem. The key insight of Cascading Decision Trees is to separate the decision path and the explanation path. Our experiments show that on average, Cascading Decision Trees generate 63.38% shorter explanation paths, avoiding overfitting and thus achieve higher test accuracy. We also empirically demonstrate that Cascading Decision Trees have advantages in the robustness against missing values.
In this paper, we propose a point cloud classification method based on graph neural network and manifold learning. Different from the conventional point cloud analysis methods, this paper uses manifold learning algorithms to embed point cloud features for better considering the geometric continuity on the surface. Then, the nature of point cloud can be acquired in low dimensional space, and after being concatenated with features in the original three-dimensional (3D)space, both the capability of feature representation and the classification network performance can be improved. We pro-pose two manifold learning modules, where one is based on locally linear embedding algorithm, and the other is a non-linear projection method based on neural network architecture. Both of them can obtain better performances than the state-of-the-art baseline. Afterwards, the graph model is constructed by using the k nearest neighbors algorithm, where the edge features are effectively aggregated for the implementation of point cloud classification. Experiments show that the proposed point cloud classification methods obtain the mean class accuracy (mA) of 90.2% and the overall accuracy (oA)of 93.2%, which reach competitive performances compared with the existing state-of-the-art related methods.
Successful motor-imagery brain-computer interface (MI-BCI) algorithms either extract a large number of handcrafted features and train a classifier, or combine feature extraction and classification within deep convolutional neural networks (CNNs). Both approaches typically result in a set of real-valued weights, that pose challenges when targeting real-time execution on tightly resource-constrained devices. We propose methods for each of these approaches that allow transforming real-valued weights to binary numbers for efficient inference. Our first method, based on sparse bipolar random projection, projects a large number of real-valued Riemannian covariance features to a binary space, where a linear SVM classifier can be learned with binary weights too. By tuning the dimension of the binary embedding, we achieve almost the same accuracy in 4-class MI ($\leq$1.27% lower) compared to models with float16 weights, yet delivering a more compact model with simpler operations to execute. Second, we propose to use memory-augmented neural networks (MANNs) for MI-BCI such that the augmented memory is binarized. Our method replaces the fully connected layer of CNNs with a binary augmented memory using bipolar random projection, or learned projection. Our experimental results on EEGNet, an already compact CNN for MI-BCI, show that it can be compressed by 1.28x at iso-accuracy using the random projection. On the other hand, using the learned projection provides 3.89% higher accuracy but increases the memory size by 28.10x.
Despite transformers' impressive accuracy, their computational cost is often prohibitive to use with limited computational resources. Most previous approaches to improve inference efficiency require a separate model for each possible computational budget. In this paper, we extend PoWER-BERT (Goyal et al., 2020) and propose Length-Adaptive Transformer that can be used for various inference scenarios after one-shot training. We train a transformer with LengthDrop, a structural variant of dropout, which stochastically determines a sequence length at each layer. We then conduct a multi-objective evolutionary search to find a length configuration that maximizes the accuracy and minimizes the efficiency metric under any given computational budget. Additionally, we significantly extend the applicability of PoWER-BERT beyond sequence-level classification into token-level classification with Drop-and-Restore process that drops word-vectors temporarily in intermediate layers and restores at the last layer if necessary. We empirically verify the utility of the proposed approach by demonstrating the superior accuracy-efficiency trade-off under various setups, including span-based question answering and text classification. Code is available at https://github.com/clovaai/length-adaptive-transformer.
An essential part for the accurate classification of electrocardiogram (ECG) signals is the extraction of informative yet general features, which are able to discriminate diseases. Cardiovascular abnormalities manifest themselves in features on different time scales: small scale morphological features, such as missing P-waves, as well as rhythmical features apparent on heart rate scales. For this reason we incorporate a variant of the complex wavelet transform, called a scatter transform, in a deep residual neural network (ResNet). The former has the advantage of being derived from theory, making it well behaved under certain transformations of the input. The latter has proven useful in ECG classification, allowing feature extraction and classification to be learned in an end-to-end manner. Through the incorporation of trainable layers in between scatter transforms, the model gains the ability to combine information from different channels, yielding more informative features for the classification task and adapting them to the specific domain. For evaluation, we submitted our model in the official phase in the PhysioNet/Computing in Cardiology Challenge 2020. Our (Team Triage) approach achieved a challenge validation score of 0.640, and full test score of 0.485, placing us 4th out of 41 in the official ranking.
We proposed a model for the Conference of Music and Technology (CSMT2020) data challenge of melody classification. Our model used the Performance Event Vector as the input sequence to build a Bidirectional RNN network for classfication. The model achieved a satisfying performance on the development dataset and Wikifonia dataset. We also discussed the effect of several hyper-parameters, and created multiple prediction outputs for the evaluation dataset.
Emotion stimulus detection is the task of finding the cause of an emotion in a textual description, similar to target or aspect detection for sentiment analysis. Previous work approached this in three ways, namely (1) as text classification into an inventory of predefined possible stimuli ("Is the stimulus category A or B?"), (2) as sequence labeling of tokens ("Which tokens describe the stimulus?"), and (3) as clause classification ("Does this clause contain the emotion stimulus?"). So far, setting (3) has been evaluated broadly on Mandarin and (2) on English, but no comparison has been performed. Therefore, we aim to answer whether clause classification or sequence labeling is better suited for emotion stimulus detection in English. To accomplish that, we propose an integrated framework which enables us to evaluate the two different approaches comparably, implement models inspired by state-of-the-art approaches in Mandarin, and test them on four English data sets from different domains. Our results show that sequence labeling is superior on three out of four datasets, in both clause-based and sequence-based evaluation. The only case in which clause classification performs better is one data set with a high density of clause annotations. Our error analysis further confirms quantitatively and qualitatively that clauses are not the appropriate stimulus unit in English.
We propose a novel method for selective classification (SC), a problem which allows a classifier to abstain from predicting some instances, thus trading off accuracy against coverage (the fraction of instances predicted). In contrast to prior gating or confidence-set based work, our proposed method optimises a collection of class-wise decoupled one-sided empirical risks, and is in essence a method for explicitly finding the largest decision sets for each class that have few false positives. This one-sided prediction (OSP) based relaxation yields an SC scheme that attains near-optimal coverage in the practically relevant high target accuracy regime, and further admits efficient implementation, leading to a flexible and principled method for SC. We theoretically derive generalization bounds for SC and OSP, and empirically we show that our scheme strongly outperforms state of the art methods in coverage at small error levels.
Deep learning methods have shown considerable potential for hyperspectral image (HSI) classification, which can achieve high accuracy compared with traditional methods. However, they often need a large number of training samples and have a lot of parameters and high computational overhead. To solve these problems, this paper proposes a new network architecture, LiteDepthwiseNet, for HSI classification. Based on 3D depthwise convolution, LiteDepthwiseNet can decompose standard convolution into depthwise convolution and pointwise convolution, which can achieve high classification performance with minimal parameters. Moreover, we remove the ReLU layer and Batch Normalization layer in the original 3D depthwise convolution, which significantly improves the overfitting phenomenon of the model on small sized datasets. In addition, focal loss is used as the loss function to improve the model's attention on difficult samples and unbalanced data, and its training performance is significantly better than that of cross-entropy loss or balanced cross-entropy loss. Experiment results on three benchmark hyperspectral datasets show that LiteDepthwiseNet achieves state-of-the-art performance with a very small number of parameters and low computational cost.
Current text classification methods typically require a good number of human-labeled documents as training data, which can be costly and difficult to obtain in real applications. Humans can perform classification without seeing any labeled examples but only based on a small set of words describing the categories to be classified. In this paper, we explore the potential of only using the label name of each class to train classification models on unlabeled data, without using any labeled documents. We use pre-trained neural language models both as general linguistic knowledge sources for category understanding and as representation learning models for document classification. Our method (1) associates semantically related words with the label names, (2) finds category-indicative words and trains the model to predict their implied categories, and (3) generalizes the model via self-training. We show that our model achieves around 90% accuracy on four benchmark datasets including topic and sentiment classification without using any labeled documents but learning from unlabeled data supervised by at most 3 words (1 in most cases) per class as the label name.
This paper proposes piece-wise matching layer as a novel layer in representation learning methods for electrocardiogram (ECG) classification. Despite the remarkable performance of representation learning methods in the analysis of time series, there are still several challenges associated with these methods ranging from the complex structures of methods, the lack of generality of solutions, the need for expert knowledge, and large-scale training datasets. We introduce the piece-wise matching layer that works based on two levels to address some of the aforementioned challenges. At the first level, a set of morphological, statistical, and frequency features and comparative forms of them are computed based on each periodic part and its neighbors. At the second level, these features are modified by predefined transformation functions based on a receptive field scenario. Several scenarios of offline processing, incremental processing, fixed sliding receptive field, and event-based triggering receptive field can be implemented based on the choice of length and mechanism of indicating the receptive field. We propose dynamic time wrapping as a mechanism that indicates a receptive field based on event triggering tactics. To evaluate the performance of this method in time series analysis, we applied the proposed layer in two publicly available datasets of PhysioNet competitions in 2015 and 2017 where the input data is ECG signal. We compared the performance of our method against a variety of known tuned methods from expert knowledge, machine learning, deep learning methods, and the combination of them. The proposed approach improves the state of the art in two known completions 2015 and 2017 around 4% and 7% correspondingly while it does not rely on in advance knowledge of the classes or the possible places of arrhythmia.
Objective: We used deep convolutional neural networks (DCNNs) to classify electroencephalography (EEG) signals in a steady-state visually evoked potentials (SSVEP) based single-channel brain-computer interface (BCI), which does not require calibration on the user. Methods: EEG signals were converted to spectrograms and served as input to train DCNNs using the transfer learning technique. We also modified and applied a data augmentation method, SpecAugment, generally employed for speech recognition. Furthermore, for comparison purposes, we classified the SSVEP dataset using Support-vector machines (SVMs) and Filter Bank canonical correlation analysis (FBCCA). Results: Excluding the evaluated user's data from the fine-tuning process, we reached 82.2% mean test accuracy and 0.825 mean F1-Score on 35 subjects from an open dataset, using a small data length (0.5 s), only one electrode (Oz) and the DCNN with transfer learning, window slicing (WS) and SpecAugment's time masks. Conclusion: The DCNN results surpassed SVM and FBCCA performances, using a single electrode and a small data length. Transfer learning provided minimal accuracy change, but made training faster. SpecAugment created a small performance improvement and was successfully combined with WS, yielding higher accuracies. Significance: We present a new methodology to solve the problem of SSVEP classification using DCNNs. We also modified a speech recognition data augmentation technique and applied it to the context of BCIs. The presented methodology surpassed performances obtained with FBCCA and SVMs (more traditional SSVEP classification methods) in BCIs with small data lengths and one electrode. This type of BCI can be used to develop small and fast systems.